Most software engineers interact with some form of code differencing every day, either directly or indirectly. Yet, many existing algorithms and tools used in that context have not significantly evolved since the basic Unix diff utility. As a consequence, many specific characteristics and semantics of code are not taken into account in the process, making the task of analysing the output of such tools difficult for developers and causing valuable time to be lost. In this paper, we describe a concrete industrial use case that could benefit from more advanced ways of performing differencing. Additionally, we provide two leads to solutions to solve the problems faced in this context.